Loading...

Web User Profile Clustering Using Artificial Immune System

Azimpour Kivi, Mozhgan | 2011

798 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: English
  3. Document No: 42240 (52)
  4. University: Sharif University of Technology, International Campus, Kish Island
  5. Department: Science and Engineering
  6. Advisor(s): Azmi, Reza; Khansari, Mohammad
  7. Abstract:
  8. Nowadays, a need for intelligent system that can assist web users in acquiring their desired information has been highlighted. In particular a mechanism for web user profiling which involves knowing the users and providing them with their preferred information is recommended to any website. To address this issue, Web Usage Mining (WUM) techniques have attracted many attentions. WUM is a kind of web mining process that tries to extract interesting usage patterns from the data that are obtained from the interaction of users with the web. The most important drawback of conventional, data mining based, WUM system is their inability to continuously learn the evolving patterns of web usage. On the other hand, many prevalent WUM approaches ignore the sequential nature of web navigations for defining the similarity between web sessions. Artificial Immune System (AIS) models are relatively new approaches that have outstanding features, such as learning, adaptation and robustness. These properties make them suitable for learning in dynamic and noisy environments such as the web. In this thesis we propose a new AIS based clustering approach that addresses weaknesses of conventional WUM systems for web usage profiling task. Moreover, new mutation process and suppression mechanism are proposed in the body of the proposed AIS. Moreover, we deploy a Danger Theory approach for filtering irrelevant web sessions prior to their presentation to the learning system. Besides, we propose using a new web sessions’ similarity measure for investigating the usage data from web access log files. In this similarity measure, in addition to the sequential nature of web navigations, the usage similarity of web sessions is taken into consideration. Different aspects of the proposed system are evaluated by applying the system on real world web sessions data for usage profiling task. The results show the ability of the system in discovering evolving usage patterns in only one pass over data. The system is also scalable and can handle noisy sets of web sessions.
  9. Keywords:
  10. Web User Profiling ; Web Session Clustering ; Session Similarity Measure ; Artificial Immune Network (AIN) ; Danger Theory

 Digital Object List

 Bookmark

No TOC